Dry etching is an important step in the manufacture of solid state electronic devices. Dry etching is a generic term that encompasses etching techniques in which gases, as opposed to liquid chemicals, are the primary etch medium. Examples of dry etching techniques used in the manufacture of solid state electronic devices are ion beam milling, plasma etching, and reactive ion etching ("RIE").
In the manufacture of solid state electronic devices, dry etching is frequently employed to remove a layer of one material in order to expose an underlying layer of another material. For example, dry etching processes are often employed to etch openings in a layer of insulating material in order to expose an underlying layer of conductive material. The desired endpoint of a dry etch process occurs when the layer being etched is completely removed. Unfortunately, layers subjected to an etching process are not always uniformly removed. Across the area of the layer being etched, the etch rate and the thickness of the layer may vary. Because of these variations, it will take longer to completely remove some regions of the layer than others. To completely remove the entire film, etching will typically have to continue even after some other regions of the layer have already been completely removed. The common practice of exposing a layer to an etching-process for longer than is necessary to etch through one region of the layer is referred to as overetching. In the manufacture of solid state electronic devices, a layer being removed by a dry etching process must be overetched to ensure that the entire layer is removed. Overetching causes a certain amount of the underlying material layer to be also removed in the dry etching process. Due to the shrinking dimensions of solid state electronic devices, it is increasingly important to minimize the amount of this collateral etching by precisely determining the endpoint of a dry etch process.
There are several methods for determining a dry etch process endpoint. A simple method consists of determining the average etch rate of the dry etch process, and then estimating the etch time needed to remove the desired amount of material. The primary disadvantage of using a pre-set etch time to estimate the endpoint is that there is no way to compensate for run-to-run fluctuations in etch rate. The etch rate may vary between runs because of variations in material properties, film thickness, or processing conditions.
Run-to-run fluctuations can be compensated for by a real-time determination of the endpoint during the course of the etching run. Two well-known, real-time methods of determining the endpoint of a dry etch process are emission spectroscopy and mass spectrometry. Both of these methods involve detecting changes in the concentration of chemical species in the etching chamber. For plasma-etching processes, a third real-time method is available in which the endpoint is determined by monitoring the direct current (DC) bias potential while holding the radio frequency (RF) power constant. Since the voltage/power dependence is related to plasma chemistry, a change in this relationship frequently occurs when the plasma contribution from the layer being etched is no longer present.
Although known real-time endpoint determination methods can compensate for run-to-run variations in etch rate, these methods present a number of potential problems. Sensitivity limitations in known methods require that a minimum amount of material be removed during the etch process. Thus none of these methods can be used if the layer being removed is very thin, or if the percentage of the surface area of the substrate exposed to the etching process is very small. Another potential disadvantage of the three methods is that they tend to average over local non-uniformities within the etching process. In other words, the known methods are incapable of making localized endpoint determinations.
Optical thickness measurement methods, such as ellipsometry and laser reflectometry, can provide real-time localized endpoint determinations. Optical methods involve determining when the endpoint of the process has been reached by measuring the thickness of the film being etched. Unfortunately, these optical methods may be adversely affected by the surface morphology of the layer being etched. Furthermore, the sensitivity of these optical methods decreases when the layer being etched is on the order of a few angstroms thick, because the optical interference effects for such thin layers are quite small.
A real-time localized endpoint detection method based on photoemission offers an alternative to the optical methods. Photoemission occurs when electrons are emitted from the surface of a material by photons with energies greater than the work function .phi. of that material. The energy of photons in a monochromatic beam of light is related to the wavelength of the light according to the following relation: EQU E=hc/.lambda.
where
E is the photon energy, typically expressed in electron-Volts (eV), PA1 h is Planck's constant, 4.14.times.10.sup.-15 eV.sec, PA1 c is the speed of light, 3.00.times.10.sup.17 nm/sec, and PA1 .lambda. is the wavelength of the light, typically expressed in nanometers (nm).
This relationship shows that the shorter the wavelength of light, the higher the photon energy. Since most materials have work functions in excess of around 3 eV, most materials will not photoemit unless they are exposed to a source of ultraviolet, or shorter, wavelength radiation. A photoemitting material produces a stream of electrons referred to as a photocurrent.
Endpoint determination techniques based on photoemission could have a number of advantages over optical endpoint determination methods. Techniques based on photoemission provide real-time, localized measurements of etch rates that are relatively insensitive to surface morphology. In addition, photoemission techniques remain sensitive even when the layer being removed is only a few angstroms thick.
Techniques involving photoemission have been utilized to characterize the steady-state properties of materials. For example, photoemission techniques have been employed to determine the doping levels in semiconductors, and to measure the thickness of contaminant films. These steady-state measurements are intended to produce an absolute value of photocurrent that can be correlated with the value of a physical property, such as thickness. Obtaining an accurate absolute value of photocurrent, however, presents a number of practical difficulties. For example, the absolute value of the photocurrent emanating from a semiconductor material is strongly affected by the presence of adsorbed gases on the surface of the semiconductor, and by photovoltaic currents that are induced in the semiconductor when the photon energy exceeds the band gap energy of the semiconductor. Most of the practical difficulties associated with obtaining an accurate absolute value of photocurrent can be avoided if all that needs to be detected is a change in a series of photocurrent measurements.